• 한국통계학회:학술대회논문집
• /
• 한국통계학회 2004년도 학술발표논문집
• /
• pp.1-8
• /
• 2004

We address the problem of parameter estimation in multivariate distributions under ignorable non-monotone missing data. The factoring likelihood method for monotone missing data, termed by Robin (1974), is extended to a more general case of non-monotone missing data. The proposed method is algebraically equivalent to the Newton-Raphson method for the observed likelihood, but avoids the burden of computing the first and the second partial derivatives of the observed likelihood Instead, the maximum likelihood estimates and their information matrices for each partition of the data set are computed separately and combined naturally using the generalized least squares method. A numerical example is also presented to illustrate the method.

• Journal of the Korean Data and Information Science Society
• /
• 제24권2호
• /
• pp.391-399
• /
• 2013

Partially linear regression is capable of providing more complete description of the linear and nonlinear relationships among random variables. In support vector regression (SVR) the hyper-parameters are known to affect the performance of regression. In this paper we propose an iterative reweighted least squares (IRWLS) procedure to solve the quadratic problem of partially linear support vector regression with a modified loss function, which enables us to use the generalized approximate cross validation function to select the hyper-parameters. Experimental results are then presented which illustrate the performance of the partially linear SVR using IRWLS procedure.

• Journal of the Korean Data and Information Science Society
• /
• 제18권3호
• /
• pp.735-744
• /
• 2007

Support vector machine(SVM) is capable of providing a more complete description of the linear and nonlinear relationships among random variables. In this paper we propose a sparse kernel regression(SKR) to overcome a weak point of SVM, which is, the steep growth of the number of support vectors with increasing the number of training data. The iterative reweighted least squares(IRWLS) procedure is used to solve the optimal problem of SKR with a Laplacian prior. Furthermore, the generalized cross validation(GCV) function is introduced to select the hyper-parameters which affect the performance of SKR. Experimental results are then presented which illustrate the performance of the proposed procedure.

• Communications for Statistical Applications and Methods
• /
• 제17권2호
• /
• pp.183-191
• /
• 2010

Support vector quantile regression(SVQR) is capable of providing more complete description of the linear and nonlinear relationships among random variables. In this paper we propose an iterative reweighted least squares(IRWLS) procedure to solve the problem of SVQR with a weighted quadratic loss function. Furthermore, we introduce the generalized approximate cross validation function to select the hyperparameters which affect the performance of SVQR. Experimental results are then presented which illustrate the performance of the IRWLS procedure for SVQR.

• The Journal of the Acoustical Society of Korea
• /
• 제15권1E호
• /
• pp.59-64
• /
• 1996

When the input-output record is available, the identification of a bilinear system is considered. It is assumed that the input is noise free and the output is contaminated by an additive noise. It is further assumed that the covariance matrix of the noise is known up to a factor of proportionality. The extended generalized total least squares (e-GTLS) method is proposed as one of the consistent estimators of the bilinear system parameters. Considering that the input is noise-free and that bilinear system equation is linear with respect to the system parameters, we extend the GTLS problem. The extended GTLS problem is reduced to an unconstrained minimization problem, and is solved by the Newton-Raphson method. We compare the GTLS method and the e-GTLS method in the point of the accuracy of the estimated system parameters.

• 수산경영론집
• /
• 제42권2호
• /
• pp.69-83
• /
• 2011

This paper estimates demand functions of oyster as Kimchi's ingredients of capital area, other areas excluding a capital area, and a whole area in Korea to forecast its demand quantities in 2011~2015. To estimate oyster demand function, this paper uses pooled data produced from Korean housewives over 30 years old in 2009 and 2010. Also, this paper adopts several econometrics methods such as Ordinary Least Squares and Feasible Generalized Least Squares. First of all, to choose appropriate variables of oyster demand functions by area, this paper carries out model's specification with joint significance test. Secondly, to remedy heteroscedasticity with pooled data, this paper attempts residual plotting between estimated squared residuals and estimated dependent variable and then, if it happens, undertakes White test to care the problem. Thirdly, to test multicollinearity between variables with pooled data, this paper checks correlations between variables by area. In this analysis, oyster demand functions of a capital area and a whole area need price of the oyster, price of the cabbage for Gimjang, and income as independent variables. The function on other areas excluding a capital area only needs price of the oyster and income as ones. In addition, the oyster demand function of a whole area needed White test to care a heteroscedasticity problem and demand functions of the other two regions did not have the problem. Thus, first model was estimated by FGLS and second two models were carried out by OLS. The results suggest that oyster demand quantities per a household as Kimchi's ingredients are going to slightly increase in a capital area and a whole area, but slightly decrease in other areas excluding a capital area in 2011~2015. Also, the results show that oyster demand quantities as kimchi's ingredients for total household targeting housewives over 30 years old are going to slightly increase in three areas in 2011~2015.

• 제어로봇시스템학회논문지
• /
• 제4권6호
• /
• pp.780-785
• /
• 1998

In this paper, it is shown that the conventional methods for dealing with the singularity problem of a manipulator can be generalized as a local minimization problem with differently weighted objective functions. A new damping method proposed in this article automatically determines the damping amounts for singular values, which are inversely proportional to the magnitude of the singular values. Furthermore, this can be done without explicitly computing the singular values. The proposed method can be applied to all the manipulators with revolute joints.

• Coupled systems mechanics
• /
• 제11권1호
• /
• pp.33-41
• /
• 2022

Quite often we have a lot of measurement data and would like to find some relation between them. One common task is to see whether some measured data or a curve of known shape fit into the cumulative measured data. The problem can be visualized since data could generally be presented as curves or planes in Cartesian coordinates where each curve could be represented as a vector. In most cases we have measured the cumulative 'curve', we know shapes of other 'curves' and would like to determine unknown coefficients that multiply the known shapes in order to match the measured cumulative 'curve'. This problem could be presented in more complex variants, e.g., a constant could be added, some missing (unknown) data vector could be added to the measured summary vector, and instead of constant factors we could have polynomials, etc. All of them could be solved with slightly extended version of the procedure presented in the sequel. Solution procedure could be devised by reformulating the problem as a measurement problem and applying the generalized inverse of the measurement matrix. Measurement problem often has some errors involved in the measurement data but the least squares method that is comprised in the formulation quite successfully addresses the problem. Numerical examples illustrate the solution procedure.

• 한국음향학회지
• /
• 제14권3호
• /
• pp.37-48
• /
• 1995

일반화된 완전최소자승법 (generalized total least squares method, GTLS)의 ARMA 시스템 식별에의 적용과 GTLS의 적응알고리듬에 대하여 논한다. 일반화된 완전최소자승법은 일별과 출력을 알고 있는 시스템식별 (system identification)문제에서, 출력이 잡음에 의하여 오염된 경우, 편이되지 않은 해를 구하기 위하여 사용되는 방법이다. 본 논문에서는 먼저 GTLS를 ARMA 시스템 식별에 적용하기 위한 formulation을 하고, 일반화된 완전최소자승법의 일반 해의 성질과 역행렬 정리 (matrix inverse lemma)를 이용하여 적응 GTLS 방법을 제안한다. 다음 제안된 방법을 통하여 시스템식별에 적용하여 그 성능을 평가한다. 또한 GTLS 알고리듬과 제안한 적응 GTLS 알고리듬의 성능을 수학적으로 해석하고 컴퓨터 시뮬레이션을 통하여 이를 검증한다.

• Journal of the Korean Data and Information Science Society
• /
• 제20권4호
• /
• pp.749-755
• /
• 2009

Support vector classification (SVC) provides more complete description of the lin-ear and nonlinear relationships between input vectors and classifiers. In this paper. we propose the sparse kernel classifier to solve the optimization problem of classification with a modified hinge loss function and absolute loss function, which provides the efficient computation and the sparsity. We also introduce the generalized cross validation function to select the hyper-parameters which affects the classification performance of the proposed method. Experimental results are then presented which illustrate the performance of the proposed procedure for classification.